Dive Club 🤿 – Geoffrey Litt: The Future of Malleable Software

Host: Ridd
Guest: Geoffrey Litt (Design Engineer, Notion; ex-Ink & Switch)
Release Date: November 14, 2025

Episode Overview

In this thoughtful episode, Ridd sits down with design engineer Geoffrey Litt to examine the future of malleable (“moldable”) software — digital products that users can shape and customize for their unique needs. They explore the impact of large language models (LLMs) and AI on software design, Geoffrey’s journey from edtech to Notion, experiments in collaborative documents and version control, and the changing role of designers. The conversation is rooted in practical examples, candid philosophies, and probing on how “document-centric computing” and AI-powered primitives could reimagine creativity and productivity software.

Key Discussion Points & Insights

1. The Coming Platform Shift with AI

• AI is driving volatility and reimagining what's possible.
  • “We're in a time of such volatility right now with AI just kind of forcing everyone to reimagine how things work... there’s actually going to be a serious platform shift over the next few years.” (Geoffrey, 00:08 & 08:01)
• Natural language as an input is powerful, but output should go beyond text.
  • “Natural language text is really good for human to AI input, especially with voice… But for the AI coming back to us, I think we can be much more creative with non-textual representations and data visualizations.” (Geoffrey, 00:24)

2. Geoffrey’s Origin Story: Why He Cares About Malleable Software

• Problem: Traditional SaaS is too rigid; real user needs are context-dependent
  • “I found it was just really, really difficult without having been in my user's shoes… Sometimes we would ask users, what do you think of our product? And they would say, oh, I actually don't use your data reports. I just export the data to Excel and do it myself… they would do exactly what that user wanted them to do.” (Geoffrey, 01:18)
• Inspiration: Let users build local, context-specific tools
  • “I couldn't get that idea out of my head: why can't more people build software tools for themselves in their local context?” (Geoffrey, 01:18)

3. Experiments at Ink & Switch: “Embark” and “Patchwork”

• Embark: Document-centric, composable travel planning
  • Centralizes context from multiple apps (weather, routes, notes) into one dynamic document
  • “Why doesn't your weather app know that you're going to be in London next week?...what we landed on was starting with just a note… then you could start pulling in bits of functionality into your note directly.” (Geoffrey, 05:16)
• Patchwork: Collaboration & branching for non-developers
  • “On any document in Patchwork, you can hit a button and get a branch… That's your private copy off in the corner that you can work on… When you feel good about your branch, you can merge it… Then we added AI bots that use all this shared branching infrastructure.” (Geoffrey, 25:05)
  • Shows how AI and human collaboration both benefit from strong version-control primitives

4. Notion as the Document-Centric Platform

• Why Notion?
  • Adopts a document-centric computing philosophy that prioritizes tasks over “build an app” paradigms
  • “Most people don't wake up in the morning wanting to build apps. People just want to get stuff done… Notion, in my opinion, is the best incarnation of document centric computing that's out there.” (Geoffrey, 09:24)
• Value comes from context, interoperability, and composable primitives
  • “The value of having all of your context in one place has never been more clear... there has to be kind of this bridging to other pieces of software.” (Geoffrey, 11:26)

5. The Philosophy of Primitives, Structure, and Simplicity

• Expose systemic structure instead of hiding complexity
  • “Good design exposes systemic structure and trusts that people will figure out how to use it.” (Geoffrey, 14:22)
  • “A lot of software looks like the nightmare bicycle to me... teaching people these very specific bits of functionality... It’s not teaching them deeper structures that they can use to do more stuff themselves.” (Geoffrey, 14:22)
• Simplicity emerges from exposing the right abstractions
  • “That's what true simplicity is, when you expose the fundamental structure and let people take advantage of it… Designing for the 40th hour of use, not the first minute.” (Geoffrey, 16:50)

6. The Tension of AI Chat, User Onboarding, and Power/Beginner Flows

• AI chat is a power-user interface but needs learnable primitives for most users
  • “AI chat is actually a very power user interface... If you're really virtuosic... you can go really far… But it's a lot harder to learn when you don't have the buttons to click.” (Geoffrey, 18:53)
  • Advocates for hybrid approaches: chat for flexibility, visualizations for feedback, and better collaborative structures

7. Canvas and Branching as Next-Gen Collaboration Metaphors

• Canvas UIs enable spatial organization and exploration
  • “It's very powerful to have a visual way to represent that thought process and see it unfold in space... This is actually a theme that I think a lot about: human-AI collaboration is fundamentally a version control problem.” (Geoffrey, 22:29)
  • Importance of making alternatives, reviewing, branching as creative process tools, not just for engineers

8. Disposable vs. Lasting Software — What Does “Malleability” Really Mean?

• Malleability is about durable, user-customized environments
  • “When I say malleable software, I do not mean only disposable software. I think of it much more as crafting an environment over time that’s actually more stable and predictable.” (Geoffrey, 30:47)
• Disposable ‘jigs’ have their place for specific, temporary tasks, aided by AI
  • “I call these tools ‘jigs’ from woodworking…a tool just for the purpose of making one project. And I think that's also a really cool pattern.” (Geoffrey, 32:21)

9. The Bottlenecks to Mainstreaming Malleable Software

• Bottlenecks are now infrastructural, not coding skills
  • “The entire way that we build and distribute software was built around an old assumption that most people can't code… When everyone's constantly editing their tools and making new tools, how do we share those?” (Geoffrey, 37:04)
  • Platforms like Notion, that combine sharing, building, and collaboration, are early answers

10. Designers in a Fluid, Malleable World

• Changing the designer’s role: From designing products to designing pattern languages
  • Inspiration from architect Christopher Alexander:
    • “The role of Chris Alexander as an architect or designer isn't to design your house, is to design a set of patterns and teach you that set of patterns, and then you go design your house.” (Geoffrey, 40:37)
  • Designers create and teach the “systems,” empowering users to combine building blocks for their context

11. Product Defaults, Templates, and Enabling User Mastery

• Defaults are important; AI can help bridge the gap to virtuosic use
  • “It's not everyone starting from scratch every time… you start somewhere that someone else already thought about, and then when you want to tweak, you can.” (Geoffrey, 44:24)
  • “Can AI help you learn about the primitives…and help teach you to actually be creative and virtuosic in arranging them in your head?” (Geoffrey, 44:24)

12. The “Coding Like a Surgeon” Mindset

• Not everyone wants to become a ‘manager of AIs’; creative control matters
  • “A surgeon does the damn surgery. They don’t sit in some admin office while someone else does the surgery... How do we think of AI as a way to leverage our time better and stay connected to the work and doing it ourselves?” (Geoffrey, 48:06)
• Flow state and intentional friction are vital in creative practice
  • “Being in flow is a really important part of getting to good work for me. I think sometimes people have this really mistaken view of the creative process, where they imagine...if you’re making ceramics, these people would think you can just say, like, oh, I want a pot that, like, looks like that. And then a robot could go make it…but no, like, when you do ceramics, you’re, like, feeling the clay…you’re shaping something and as it takes shape, you’re responding.” (Geoffrey, 51:07)

13. Learning With and From AI: The Designer’s Path Forward

• See AI as a learning partner, not just an automation tool
  • “We have the best learning machines ever invented now, but they happen to also be cheat on your homework machines at the same time. And so every day we get to make a choice. Are we going to cheat on our homework? Or are we going to get better?” (Geoffrey, 53:03)
• AI could help explain, not just generate; demand better “explanatory UX”
  • “Why aren't we doing that? I mean…can we do better? Maybe it just goes and does the whole thing and then it comes back to you with some learning experience…with interactive diagrams and…quiz questions.” (Geoffrey, 54:35)
  • “AI can help us be more critical with our thinking, not less.” (Geoffrey, 55:34)

Notable Quotes & Memorable Moments

• On Document-Centric AI:
  “When you plan a trip in Notion, what if somehow that document came alive with all the tools you need to plan that trip?” (Geoffrey, 09:24)
• On Simplicity:
  “I think that's what true simplicity is, when you expose the fundamental structure and let people take advantage of it.” (Geoffrey, 16:50)
• On User Agency:
  “One of the reasons I care about malleability is that I think it cultivates a mindset for people. It's a mindset of agency, of believing that you can change your environments and your surroundings.” (Geoffrey, 32:21)
• On Defaults & Teaching Primitives:
  “Ultimately the really best tools are only going to come if you actually are becoming fluent in the primitives.” (Geoffrey, 44:24)
• On AI as a Learning Partner:
  “Every day we get to make a choice. Are we going to cheat on our homework? Or are we going to get better?” (Geoffrey, 53:03)
• On the Future of Designers:
  “There is a role for design expertise, but it's more about designing the system and then teaching the user to do it, the actual design themselves.” (Geoffrey, 40:37)

Key Timestamps

• 00:08 - AI-driven volatility and platform shift
• 01:18 - Geoffrey’s origin story: SaaS rigidity & user needs
• 05:16 - Ink & Switch’s “Embark” and document-centric prototyping
• 09:24 - Why Notion exemplifies document-centric computing
• 14:22 - Exposing structure; “Nightmare Bicycle” paradigm
• 18:53 - The power and challenge of AI chat interfaces
• 22:29 - Canvas UIs as next-gen collaboration tools
• 25:05 - Patchwork: branching/version control for everyone
• 30:47 - Malleable does not mean disposable
• 37:04 - Bottlenecks to widespread malleable software
• 40:37 - Designer as pattern creator/educator (Christopher Alexander)
• 48:06 - Coding like a surgeon; AI as a creative assistant
• 53:03 - AI for learning vs. cheating; lessons for designers

Lessons & Advice for Designers

• Embrace learning from and with AI; don't use it just to "skip the hard parts"
• Seek ways to learn the underlying system primitives — AI can help you, but you must be intentional
• Recognize the value of designing extensible, understandable building blocks and teaching users how to adapt them
• Advocate for explanatory UX and tools that teach ("show what you did, and test my understanding," 54:35)
• Find your own pace and balance between defaults/templates and full-on customization as you level up

Closing Thoughts

Geoffrey Litt paints a compelling vision for software where the line between builder and user blurs — with tools that are both deeply composable and approachable, AI-powered and yet agency-enhancing. The future, in his eyes, is not about less friction, but the right kind of friction — enabling flow, mastery, and new creative frontiers for designers and non-designers alike.